Filling in the gaps—floating-point numbers